On ring extensions pinched at the integral closure (Q6981614)

Let \(R\) be an integral domain with quotient field \(\operatorname{qf}(R)\). An \textit{overring} of \(R\) is any ring \(T\) with \(R \subseteq T \subseteq \operatorname{qf}(R)\). An overring \(T\) is \textit{minimal} over \(R\) if there is no intermediate ring strictly between \(R\) and \(T\), and \(R\) is said to have a \textit{unique minimal overring} if there is exactly one such \(T \ne R\). A ring extension \(R \subseteq S\) is called a \textit{\(\lambda\)-extension} if every intermediate ring in \([R,S]\) is totally ordered by inclusion. Dually, a proper overring \(R_{1}\) of \(R\) is the \textit{unique maximal overring} of \(R\) if \(R_{1} \neq \operatorname{qf}(R)\) and every overring of \(R\) is contained in \(R_{1}\). An extension \(R \subseteq T\) is an \textit{i-extension} if the canonical map \(\operatorname{Spec}(T) \to \operatorname{Spec}(R)\) is injective, and \(R\) is an \textit{i-domain} if this holds for every overring \(T\) of \(R\).\N\NThe article first shows that the existence of a unique maximal overring is equivalent to being local, having the \(i\)-domain property, and sharing one-dimensional Krull structure with the integral closure. More precisely, let \(R\) be an integral domain that is not a field, and let \(K = \operatorname{qf}(R)\). If \(\overline{R}\) denotes the integral closure of \(R\) in \(K\), then \(\overline{R}\) is the unique maximal overring of \(R\) if and only if \(R\) is a local \(i\)-domain and\N\[\N\dim R = \dim \overline{R} = 1.\N\]\NWithin the class of Prüfer domains, an equivalent condition is that \(R\) contains a nonzero minimal divided prime ideal.\N\NA related criterion describes when the lattice \([R,S]\) of intermediate rings of an extension \(R \subset S\) splits around the integral closure \(\overline{R}\), under the assumptions that the extension has the Finite Chain Property and \(R \subset \overline{R} \subset S\). This decomposition is shown to be equivalent to a condition on minimal extensions below \(\overline{R}\), and it characterizes precisely those intermediate extensions \(J \subseteq S\) for which\N\[\N\bigl|\mathrm{M\text{Supp}}_{J}(S/J)\bigr| = 1.\N\]\NAs an application, the article shows that, for extensions with the Finite Maximal Chain (FMC) property, having a single maximal support determines when an extension is a \(\lambda\)-extension. In particular, an FMC extension \(R \subset S\) is a \(\lambda\)-extension if and only if \(\mathrm{M\text{Supp}}_{R}(S/R) = \{M\}\) and the localization \(R_{M} \subset S_{M}\) is itself a \(\lambda\)-extension. This provides a converse to the earlier observation that FMC \(\lambda\)-extensions have a unique maximal support. In the integrally closed setting, the \(\lambda\)-extension property is then determined entirely by the support: \(R \subset S\) is a \(\lambda\)-extension precisely when \(\text{Supp}_{R}(S/R)\) is totally ordered by inclusion, or equivalently, when there is a single maximal supporting ideal (i.e.\ \(\lvert \mathrm{M\text{Supp}}_{R}(S/R) \rvert = 1\)).\N\NAs one of the concluding contributions, the article also extends a structural result originally proved for \(\lambda\)-extensions. Gilbert showed that if \(R \subset T\) is a \(\lambda\)-extension, then \(T\) can have exactly two nonzero factors in any direct product decomposition. This condition is now relaxed and generalized: even without assuming the \(\lambda\)-extension property, the same two-factor restriction still holds provided that \(R \subset T\) has an overring comparable with all intermediate rings.